Knowledge Base Retrieval for Graduate Education Process Management: Trace Exceptions and Review Results

Graduate Education: review retrieval for exception review. Check source versions, input mappings and the values received by downstream nodes.

What this category's data looks like

Graduate training process management covers materials such as training plans, thesis proposals, and mid-term assessment forms, originating from supervisors, schools, and graduate management positions, and updated by semester and milestone. A training plan contains course names, credits, start and end dates, and supervisor signatures; a thesis proposal contains the title, number of references, review date, and expert signatures; a mid-term assessment form contains the assessment conclusion, date, and assessment panel signatures. When these three types of materials enter the knowledge base retrieval and recall function, each serves as a source of candidate chunks: the training plan provides course and credit fields, the thesis proposal provides topic selection and review records, and the mid-term assessment form provides stage conclusions and signature information. The names, quantities, and dates of the same record across the two types of materials must be compared item by item, and versions and signature receipts must also be verified.

Configuration and implementation checks

This step is executed by the knowledge base retrieval service (retrieval node and search test), outputting candidate chunks and similarity scores (after pattern, merge, rerank, filter, and limit trimming). The exception list and review conclusions are completed by subsequent nodes or manually.

ObjectCheckVerification action
Training planWhether course names, credits, start and end dates, and supervisor signatures are consistent with the thesis proposal and mid-term assessment formIn the search test, switch searchMode to semantic, full-text, or hybrid, compare candidate chunks and similarity scores, and check course fields and signatures
Thesis proposalWhether the title, number of references, review date, and expert signatures correspond to the training plan and mid-term assessment formTurn usingReRank on and off, compare the order of the top candidate chunks, and check whether title and date fields are consistent across materials
Mid-term assessment formWhether the assessment conclusion, date, and assessment panel signatures match the training plan and thesis proposalRecord the number of candidate chunks and similarity scores, check the assessment conclusion and signatures, and confirm whether versions are out of sync or receipts are missing
Citation limitWhether key original chunks are still retained after trimming; record the array count and the citation text tokens separatelyCheck limit and the citation text tokens, record the array count separately, and do not compare the two for equality

Three common errors

  • Symptom: the same course name is written differently in the training plan and the mid-term assessment form. Cause: field conventions were not unified at entry time, so candidate chunks cannot match names across materials.
  • Symptom: the thesis proposal date and the mid-term assessment form date differ greatly. Cause: versions are out of sync, and old materials remain in the knowledge base and participate in recall.
  • Symptom: candidate chunks lack supervisor or assessment panel signatures. Cause: signature receipts were not stored together with the materials, so retrieval results cannot support review conclusions.

How to verify the setup

Acceptance of this function: test once each with synonym-rewritten questions, compare candidate chunks and similarity scores, then turn usingReRank on and off to see the ordering of the top chunks, and switch searchMode to observe changes in candidates.

Business test: create an exception record, see whether the source material can be traced, confirm that the exception is supported by original records and reviewer signatures, and then hand it to the responsible person for handling.

  • Compare the fields, dates, and signatures of the training plan, thesis proposal, and mid-term assessment form item by item, and mark records whose names, quantities, or dates do not match.
  • Check versions and receipts, confirm materials missing signatures or receipts, and attach the candidate chunks together with the original records after the exception list.